To make the trust management system robust against sleeper attack  , where a peer behaves honestly for a sufficiently long time to acquire a good reputation and then starts misbehaving and exploiting the system, the proposed system assigns more weights to the recent observations for computing the aggregate reputation metrics of a peer. As mentioned earlier in this section, the trust value of a peer is computed as the statistical expected value of its reputation.
The trust value of a peer lies in the interval [0, 1]. In the implementation of the proposed protocol, we have used an LRU least recently used data structure which is maintained in each peer to keep track of the most recent transactions the peer had with maximum of 32 peers. However, the choice of the number of peers whose transaction history is maintained in each peer is a tuneable parameter, which can be increased or decreased based on the memory and the computing capabilities of the peers.
The public key serves as the identity of the peer. The identities are persistent and they enable two peers that have exchanged keys to locate and connect to one another whenever the peers are online. The DHT entries for the peer i are signed by the peer i and encrypted using its public key.
Each entry is indexed by a 20 byte randomly generated shared secret, which is agreed upon during the first successful connection between the two peers. Node churning model: In P2P networks, a large number of peers may join and leave at any time. This activity is termed as node churning. To simulate node churning, prior to each generation a set of consecutive searches , a fixed percentage of nodes are chosen randomly as inactive peers. These peers neither initiate nor respond to a query in that generation, and they join the system later with their LRU data structure cleared.
The clearing of the LRU data structure ensures that these peers do not have any historical information about their past transactions with other peers in the network. Since in a real world network, even in presence of churning, the approximate distribution of content categories and files remain constant, the contents of the peers undergoing churn are exchanged with the peer remaining in the network, so that the content distribution model of the network remains unchanged.
Threat model: The malicious peers adopt various strategies threat models to conceal their behavior so that they can effectively disrupt the activities in the network, and yet go undetected. The proposed protocol considers two threat models. The peers which share good quality files enjoy better topological positions after topology adaptation. In the threat model A, the malicious peers attempt to circumvent this effect by providing good files occasionally with a probability - known as degree of deception- to lure other peers to form communities with them.
In the threat model B, a group of malicious peer joins the system and provides good files until the connectivity of the peers reaches a maximum value - the edge limit. The peers then start acting maliciously by spreading fake contents in the network. An ideal search protocol should satisfy several requirements such as: a It should have a high search efficiency and search quality - i. The proposed search protocol has been designed to satisfy each of these requirements. At each peer, the query is forwarded to a subset of its neighbors; the number of neighbors is decided based on the local estimate of connectivity.
The search process operates in two steps: query initiation and query forward. These steps are described in the following. Query initiation: The initiating peer forms a query packet containing the name of the file c, r and forwards it to some of its neighbors along with the Probcom and the TTL values. The query is disseminated using the following neighbor selection rule. The neighbors are ranked based on both their trustworthiness and their similarities of interest. Preference is given to the trusted neighbors sharing similar contents. Among the trusted neighbors, the community members having their contents matched to the query are preferred.
Peer-To-Peer Computing for Mobile Networks : Information Discovery and Dissemination - brunmiddphati.tk
If the number of community links is not adequate enough, the query is forwarded through the connectivity links also. The various cases of neighbor selection are illustrated in Fig. It is assumed that in each case only two neighbors are selected for forwarding a query. When the query c2, f4 reaches the peer P, following four cases may occur. In Case 1, the peer P has sufficient number of community neighbors two community neighbors sharing files in the category c2. Hence, these peers are chosen for forwarding the query.
In this scenario, the community neighbors sharing the c2 and the c6 categories of files are preferred over the connectivity neighbor sharing the file category c2 for forwarding the query. This is because of the fact that the peers forward queries to the community peers which have higher trust values than the connectivity peers. In Case 3, there is only one community neighbor that shares the file category c2. Hence that neighbor is chosen for the purpose of query forwarding. Among the remaining connectivity neighbors, the most trusted one containing the c6 category is selected.
Articles in Refereed Conferences
In Case 4, there are no community neighbors. Assuming that the peer P has the same level of trust for all its neighbors, the neighbor sharing the matching content category c2 is chosen for forwarding the query.
Among the rest of the neighbors, the peer c6 is chosen randomly since only two forwarding peers are to be selected. Query forwarding: i Check the trust level of the peer j: The peer i checks the trust rating of the peer j through the check trust rating algorithm explained later in this section. The selection of the peers for further forwarding of the query is done accordingly. If the TTL value has not expired, the following steps are executed.
The search process is shown in Fig. It is assumed that from each peer, the query is forwarded to two neighbors. The matching community links are preferred over the connectivity links to dispatch the query. The peer 1 initiates the query and forwards it to two community neighbors 3 and 4. The query reaches the peer 8 via the peer 4. However, the peer 8 knows from its previous transactions with the peer 4 that the peer 4 is malicious.
Signal-Strength Statistical Fingerprinting-based Positioning in Wireless LANs by Maria Papadopouli
Hence, it blocks the query. The query forwarded by the peer 5 is also blocked by the peer 10 and the peer 11 as both of them know that the peer 5 is malicious. The query is matched at four peers: 4, 6, 9 and Topology Adaptation: The responses are sorted by the initiating peer i based on the reputations of the resource providers, and the peer having the highest reputation is selected as the source for downloading.
- ICS - TNL - Mobile Computing Activity - Publications.
- Signal Processing for Neuroscientists, A Companion Volume: Advanced Topics, Nonlinear Techniques and Multi-Channel Analysis (Elsevier Insights).
- Customer Reviews.
- The gods of the Egyptians; or, Studies in Egyptian mythology.
The requesting peer checks the authenticity of the downloaded file. If the file is found to be fake, the peer i attempts to download the file from other sources until it is able to find the authentic resource or it does have any sources left for searching. The peer then updates the trust ratings and possibly adapts the network topology after a failed or a successful download, to bring the trusted peers closer to its neighborhood, and to drop the malicious peers from its community.
The restructuring of the network is controlled by a parameter known as degree of rewiring which represents the probability with which a link is formed between a pair of peers. This parameter allows the trust information to propagate through the network. If the peer j finds that the peer i is malicious i. In the example shown in Fig. It reduces the trust score of the peer 4 and deletes the community link It then downloads the file from the peer 6. The peer 1 now sends a request to the peer 6, and the latter grants the request after checking its trust value.
Hence, the community edge is added.
The malicious peer 4 loses one community link and the peer 6 gains one community edge. However, the network still remains connected by the connectivity edges which are shown in dotted lines in Fig. Checking of the Trust Rating of the Peers: The trust rating of the peers is used at various stages of execution of the protocol to make a decision on the possible source for downloading a file, to stop a query forwarded from a malicious node and to adapt the topology.
When no transaction history is available, a peer seeks for the recommendations from its neighbors using a trust query message. If none of its community neighbors possesses any information about the peer j, then the peer i initiates a directed DFS search. Protection of the identity of the requesting peer: In Fig.
Once the query propagation module successfully identifies the possible supplier of the resource, the trusted peer serves as a proxy to deliver the data to the requester peer. Other peers including the supplier of the resource will not be able to know the real requester.